US10474930B1 - Learning method and testing method for monitoring blind spot of vehicle, and learning device and testing device using the same - Google Patents

Learning method and testing method for monitoring blind spot of vehicle, and learning device and testing device using the same Download PDF

Info

Publication number
US10474930B1
US10474930B1 US16/152,699 US201816152699A US10474930B1 US 10474930 B1 US10474930 B1 US 10474930B1 US 201816152699 A US201816152699 A US 201816152699A US 10474930 B1 US10474930 B1 US 10474930B1
Authority
US
United States
Prior art keywords
vehicle
testing
information
monitoring
layer
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
US16/152,699
Other languages
English (en)
Inventor
Kye-hyeon KIM
Yongjoong Kim
Insu Kim
Hak-Kyoung Kim
Woonhyun Nam
Sukhoon Boo
Myungchul Sung
Donghun Yeo
Wooju RYU
Taewoong Jang
Kyungjoong Jeong
Hongmo Je
Hojin Cho
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Stradvision Inc
Original Assignee
Stradvision Inc
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Stradvision Inc filed Critical Stradvision Inc
Priority to US16/152,699 priority Critical patent/US10474930B1/en
Assigned to StradVision, Inc. reassignment StradVision, Inc. ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: BOO, SUKHOON, CHO, HOJIN, JANG, TAEWOONG, JE, HONGMO, JEONG, KYUNGJOONG, KIM, HAK-KYOUNG, KIM, INSU, KIM, Kye-hyeon, KIM, YONGJOONG, NAM, WOONHYUN, RYU, Wooju, SUNG, MYUNGCHUL, YEO, DONGHUN
Priority to KR1020190105496A priority patent/KR102349953B1/ko
Priority to EP19195517.8A priority patent/EP3633558A1/en
Priority to JP2019182005A priority patent/JP6814334B2/ja
Priority to CN201910949146.3A priority patent/CN111008553B/zh
Application granted granted Critical
Publication of US10474930B1 publication Critical patent/US10474930B1/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G06K9/6262—
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/7715—Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W40/00—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models
    • B60W40/02—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models related to ambient conditions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
    • G06K9/00791—
    • G06K9/4671—
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
    • G06V10/449—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
    • G06V10/451—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters with interaction between the filter responses, e.g. cortical complex cells
    • G06V10/454—Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
    • G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
    • G06V10/464—Salient features, e.g. scale invariant feature transforms [SIFT] using a plurality of salient features, e.g. bag-of-words [BoW] representations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/776—Validation; Performance evaluation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
    • G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/16—Anti-collision systems
    • G08G1/167—Driving aids for lane monitoring, lane changing, e.g. blind spot detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/20—Ensemble learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]

Definitions

  • the present invention relates to a learning method and a testing method for monitoring one or more blind spots of one or more vehicles, and a learning device and a testing device using the same; and more particularly, to the learning method of a CNN (Convolutional Neural Network) for monitoring one or more blind spots of a monitoring vehicle, including steps of: (a) a learning device, if a training image corresponding to at least one video image taken from the monitoring vehicle is inputted, instructing a detector on the monitoring vehicle to output class information and location information on a monitored vehicle included in the training image; (b) the learning device instructing a cue information extracting layer to perform one or more operations by using the class information and the location information on the monitored vehicle, to thereby output one or more pieces of cue information on the monitored vehicle, and instructing an FC layer for monitoring the blind spots to perform one or more neural network operations by using said pieces of cue information or their processed values on the monitored vehicle, to thereby output a result of determining whether the monitored vehicle is located on one of the blind spots of the monitoring vehicle; and
  • a monitoring vehicle has a side view mirror for each side thereof and a rear-view mirror at the front center of its cabin for a good field of view of the side and the rear needed for change of lanes by a driver.
  • the side view mirror is used for seeing each side and the rear thereof, it has a blind spot (BS) where the driver cannot see a monitored vehicle or any other objects that are very close thereto.
  • BS blind spot
  • the driver sometimes put a convex mirror onto a corner of the side view mirror, which enables the driver to see the blind spot.
  • the driver must see the blind spot with his/her own eyes to change lanes which puts further strain to the driver, and there may exist the blind spot that still cannot be seen through the convex mirror even if the driver alters his/her head position.
  • a conventional blind spot monitoring system aims to prevent accidents from happening when the driver changes lanes without noticing the monitored vehicle in the blind spot, by providing the driver with information on a detection of the monitored vehicle, that is located in the blind spot or approaching the blind spot, through a sensor placed at the rear of the monitoring vehicle.
  • the conventional blind spot monitoring system using a vision sensor may detect monitored vehicles included in video information and determine whether the monitored vehicles are on the blind spots by using information on the detected monitored vehicles.
  • the conventional blind spot monitoring system using the vision sensor may need some logic to determine whether the monitored vehicles are located on the blind spots by using output signals from a detector which detects the monitored vehicles in the video information.
  • the conventional blind spot monitoring system using the vision sensor may have a problem that an appropriate logic capable of determining whether the monitored vehicles are on the blind spots has to be designed depending on the detector to be used.
  • the conventional blind spot monitoring system may have another problem of requiring lots of time to develop because the appropriate logic capable of determining whether the monitored vehicles are on the blind spots has to be designed depending on output characteristics of the designed detector.
  • a learning method of a CNN (Convolutional Neural Network) for monitoring one or more blind spots of a monitoring vehicle including steps of: (a) a learning device, if a training image corresponding to at least one video image taken from the monitoring vehicle is inputted, instructing a detector on the monitoring vehicle to output class information and location information on a monitored vehicle included in the training image; (b) the learning device instructing a cue information extracting layer to perform one or more operations by using the class information and the location information on the monitored vehicle, to thereby output one or more pieces of cue information on the monitored vehicle, and instructing an FC layer for monitoring the blind spots to perform one or more neural network operations by using said pieces of cue information or their processed values on the monitored vehicle, to thereby output a result of determining whether the monitored vehicle is located on one of the blind spots of the monitoring vehicle; and (c) the learning device instructing a first loss layer to generate one or more loss values for the blind spots by referring to the result and its
  • the detector is a vehicle detector based on an R-CNN (Region-based Convolutional Neural Network) including: one or more convolutional layers which generate a feature map from the training image, an RPN (Region Proposal Network) which generates an ROI (Region Of Interest) of the monitored vehicle from the feature map, a pooling layer which generates a feature vector by pooling an area, in the feature map, corresponding to the ROI, at least one FC layer for the vehicle detection which performs at least one fully connected operation on the feature vector, to thereby generate one or more FC output values, a classification layer which outputs the class information on the monitored vehicle by referring to the FC output values, and a regression layer which outputs the location information on the monitored vehicle by referring to the FC output values.
  • R-CNN Registered-based Convolutional Neural Network
  • the loss values for the vehicle detection include one or more class loss values and one or more location loss values of the monitored vehicle, wherein the learning device learns one or more parameters of the FC layer for the vehicle detection and one or more parameters of the convolutional layers by backpropagating the class loss values and the location loss values.
  • the learning device instructs the CNN to receive a concatenated value acquired by using said pieces of cue information on the monitored vehicle generated by the cue information extracting layer and the feature vector generated by the pooling layer of the detector, as an input to the FC layer for monitoring the blind spots.
  • the FC layer for monitoring the blind spots includes a neural network capable of outputting a result value of whether the monitored vehicle is located on one of the blind spots by a multilayer perceptron to which said pieces of cue information on the monitored vehicle are inputted.
  • said pieces of cue information on the monitored vehicle include at least part of (i) the class information on the monitored vehicle, (ii) the location information on the monitored vehicle, (iii) size information on the monitored vehicle corresponding to an ROI (Region Of Interest) size, (iv) aspect ratio information on the monitored vehicle, and (v) distance information between a center of the monitored vehicle and one of outer sides of the blind spots.
  • the cue information extracting layer determines a smaller value among either of (i) a distance of one outer boundary of the blind spot BS-L from the center of the monitored vehicle and (ii) a distance of the other outer boundary of the blind spot BS-R from the center thereof as the distance information between the center of the monitored vehicle and said one of the outer sides of the blind spots, and further outputs relative location information in addition to the distance information in order to determine whether the center of the monitored vehicle is located inside or outside of said one of the outer sides of the blind spots.
  • a testing method of a CNN for monitoring one or more blind spots of a monitoring vehicle, including steps of: (a) on condition that a learning device, (i) has instructed a detector on the monitoring vehicle to output class information for training and location information for training on a monitored vehicle included in a training image corresponding to at least one video image taken from the monitoring vehicle, (ii) has instructed a cue information extracting layer to perform one or more operations by using the class information for training and the location information for training on the monitored vehicle, to thereby output one or more pieces of cue information for training on the monitored vehicle, (iii) has instructed an FC layer for monitoring the blind spots to perform one or more neural network operations by using said pieces of cue information for training or their processed values, to thereby output a result for training of determining whether the monitored vehicle is located on one of the blind spots, and (iv) has instructed a first loss layer to generate one or more loss values for the blind spots by referring to the result
  • the detector is a vehicle detector based on an R-CNN (Region-based Convolutional Neural Network) including: one or more convolutional layers which generate a feature map for testing from the test image, an RPN (Region Proposal Network) which generates an ROI (Region Of Interest) for testing of the monitored vehicle from the feature map for testing, a pooling layer which generates a feature vector for testing by pooling an area, in the feature map for testing, corresponding to the ROI for testing, at least one FC layer for the vehicle detection which performs at least one fully connected operation on the feature vector for testing, to thereby generate one or more FC output values, a classification layer which outputs the class information for testing on the monitored vehicle by referring to the FC output values, and a regression layer which outputs the location information for testing on the monitored vehicle by referring to the FC output values.
  • R-CNN Registered-based Convolutional Neural Network
  • the testing device instructs the CNN to receive a concatenated value for testing acquired by using said pieces of cue information for testing on the monitored vehicle generated by the cue information extracting layer and the feature vector for testing generated by the pooling layer of the detector, as an input to the FC layer for monitoring the blind spots.
  • the FC layer for monitoring the blind spots includes a neural network capable of outputting a result value for testing of whether the monitored vehicle is located on one of the blind spots by a multilayer perceptron to which said pieces of cue information for testing or their processed values on the monitored vehicle are inputted.
  • said pieces of cue information for testing on the monitored vehicle include at least part of (i) the class information for testing on the monitored vehicle, (ii) the location information for testing on the monitored vehicle, (iii) size information for testing on the monitored vehicle corresponding to an ROI size, (iv) aspect ratio information for testing on the monitored vehicle, and (v) distance information for testing between a center of the monitored vehicle and one of outer sides of the blind spots.
  • the cue information extracting layer determines a smaller value among either of (i) a distance of one outer boundary of the blind spot BS-L from the center of the monitored vehicle and (ii) a distance of the other outer boundary of the blind spot BS-R from the center thereof as the distance information for testing between the center of the monitored vehicle and said one of the outer sides of the blind spots, and further outputs relative location information for testing in addition to the distance information for testing in order to determine whether the center of the monitored vehicle is located inside or outside of said one of the outer sides of the blind spots.
  • the monitoring vehicle for the testing device is not identical to the monitoring vehicle for the learning device.
  • a learning device of a CNN for monitoring one or more blind spots of a monitoring vehicle, including: a communication part for receiving a training image corresponding to at least one video image taken from the monitoring vehicle; and a processor for (I) instructing a detector on the monitoring vehicle to output class information and location information on a monitored vehicle included in the training image, (II) instructing a cue information extracting layer to perform one or more operations by using the class information and the location information on the monitored vehicle, to thereby output one or more pieces of cue information on the monitored vehicle, and instructing an FC layer for monitoring the blind spots to perform one or more neural network operations by using said pieces of cue information or their processed values on the monitored vehicle, to thereby output a result of determining whether the monitored vehicle is located on one of the blind spots of the monitoring vehicle, and (III) instructing a first loss layer to generate one or more loss values for the blind spots by referring to the result and its corresponding first GT (G)
  • the detector is a vehicle detector based on an R-CNN (Region-based Convolutional Neural Network) including: one or more convolutional layers which generate a feature map from the training image, an RPN (Region Proposal Network) which generates an ROI (Region Of Interest) of the monitored vehicle from the feature map, a pooling layer which generates a feature vector by pooling an area, in the feature map, corresponding to the ROI, at least one FC layer for the vehicle detection which performs at least one fully connected operation on the feature vector, to thereby generate one or more FC output values, a classification layer which outputs the class information on the monitored vehicle by referring to the FC output values, and a regression layer which outputs the location information on the monitored vehicle by referring to the FC output values.
  • R-CNN Registered-based Convolutional Neural Network
  • the loss values for the vehicle detection include one or more class loss values and one or more location loss values of the monitored vehicle, wherein the processor learns one or more parameters of the FC layer for the vehicle detection and one or more parameters of the convolutional layers by backpropagating the class loss values and the location loss values.
  • the processor instructs the CNN to receive a concatenated value acquired by using said pieces of cue information on the monitored vehicle generated by the cue information extracting layer and the feature vector generated by the pooling layer of the detector, as an input to the FC layer for monitoring the blind spots.
  • the FC layer for monitoring the blind spots includes a neural network capable of outputting a result value of whether the monitored vehicle is located on one of the blind spots by a multilayer perceptron to which said pieces of cue information on the monitored vehicle are inputted.
  • said pieces of cue information on the monitored vehicle include at least part of (i) the class information on the monitored vehicle, (ii) the location information on the monitored vehicle, (iii) size information on the monitored vehicle corresponding to an ROI (Region Of Interest) size, (iv) aspect ratio information on the monitored vehicle, and (v) distance information between a center of the monitored vehicle and one of outer sides of the blind spots.
  • the cue information extracting layer determines a smaller value among either of (i) a distance of one outer boundary of the blind spot BS-L from the center of the monitored vehicle and (ii) a distance of the other outer boundary of the blind spot BS-R from the center thereof as the distance information between the center of the monitored vehicle and said one of the outer sides of the blind spots, and further outputs relative location information in addition to the distance information in order to determine whether the center of the monitored vehicle is located inside or outside of said one of the outer sides of the blind spots.
  • a testing device of a CNN for monitoring one or more blind spots of a monitoring vehicle, including: a communication part for, on condition that a learning device, (i) has instructed a detector on the monitoring vehicle to output class information for training and location information for training on a monitored vehicle included in a training image corresponding to at least one video image taken from the monitoring vehicle, (ii) has instructed a cue information extracting layer to perform one or more operations by using the class information for training and the location information for training on the monitored vehicle, to thereby output one or more pieces of cue information for training on the monitored vehicle, (iii) has instructed an FC layer for monitoring the blind spots to perform one or more neural network operations by using said pieces of cue information for training or their processed values, to thereby output a result for training of determining whether the monitored vehicle is located on one of the blind spots, and (iv) has instructed a first loss layer to generate one or more loss values for the blind spots by referring
  • the detector is a vehicle detector based on an R-CNN (Region-based Convolutional Neural Network) including: one or more convolutional layers which generate a feature map for testing from the test image, an RPN (Region Proposal Network) which generates an ROI (Region Of Interest) for testing of the monitored vehicle from the feature map for testing, a pooling layer which generates a feature vector for testing by pooling an area, in the feature map for testing, corresponding to the ROI for testing, at least one FC layer for the vehicle detection which performs at least one fully connected operation on the feature vector for testing, to thereby generate one or more FC output values, a classification layer which outputs the class information for testing on the monitored vehicle by referring to the FC output values, and a regression layer which outputs the location information for testing on the monitored vehicle by referring to the FC output values.
  • R-CNN Registered-based Convolutional Neural Network
  • the processor instructs the CNN to receive a concatenated value for testing acquired by using said pieces of cue information for testing on the monitored vehicle generated by the cue information extracting layer and the feature vector for testing generated by the pooling layer of the detector, as an input to the FC layer for monitoring the blind spots.
  • the FC layer for monitoring the blind spots includes a neural network capable of outputting a result value for testing of whether the monitored vehicle is located on one of the blind spots by a multilayer perceptron to which said pieces of cue information for testing or their processed values on the monitored vehicle are inputted.
  • said pieces of cue information for testing on the monitored vehicle include at least part of (i) the class information for testing on the monitored vehicle, (ii) the location information for testing on the monitored vehicle, (iii) size information for testing on the monitored vehicle corresponding to an ROI size, (iv) aspect ratio information for testing on the monitored vehicle, and (v) distance information for testing between a center of the monitored vehicle and one of outer sides of the blind spots.
  • the cue information extracting layer determines a smaller value among either of (i) a distance of one outer boundary of the blind spot BS-L from the center of the monitored vehicle and (ii) a distance of the other outer boundary of the blind spot BS-R from the center thereof as the distance information for testing between the center of the monitored vehicle and said one of the outer sides of the blind spots, and further outputs relative location information for testing in addition to the distance information for testing in order to determine whether the center of the monitored vehicle is located inside or outside of said one of the outer sides of the blind spots.
  • the monitoring vehicle for the testing device is not identical to the monitoring vehicle for the learning device.
  • FIG. 1 is a drawing schematically illustrating a learning device for monitoring blind spots of a monitoring vehicle in accordance with one example embodiment of the present invention.
  • FIG. 2 is a drawing schematically illustrating a learning method for monitoring the blind spots in accordance with one example embodiment of the present invention.
  • FIG. 3 is a drawing schematically illustrating location information on a monitored vehicle among one or more pieces of cue information on the monitored vehicle in the learning method for monitoring the blind spots in accordance with one example embodiment of the present invention.
  • FIG. 4 is a drawing schematically illustrating distance information between a center of the monitored vehicle and one of outer sides of the blind spots among said pieces of cue information on the monitored vehicle in the learning method for monitoring the blind spots in accordance with one example embodiment of the present invention.
  • FIG. 5 is a drawing schematically illustrating an FC layer for monitoring the blind spots in the learning method for monitoring the blind spots in accordance with one example embodiment of the present invention.
  • FIG. 6 is a drawing schematically illustrating a testing device for monitoring the blind spots in accordance with one example embodiment of the present invention.
  • FIG. 7 is a drawing schematically illustrating a testing method for monitoring the blind spots in accordance with one example embodiment of the present invention.
  • FIG. 8 is a diagram schematically illustrating a detector used in the testing device for monitoring the blind spots in accordance with one example embodiment of the present invention.
  • FIG. 9 is a drawing schematically illustrating a learning device for monitoring blind spots of a monitoring vehicle in accordance with another example embodiment of the present invention.
  • FIG. 10 is a diagram schematically illustrating a learning method for monitoring the blind spots in accordance with another example embodiment of the present invention.
  • FIG. 11 is a drawing schematically illustrating a testing device for monitoring the blind spots in accordance with another example embodiment of the present invention.
  • FIG. 12 is a diagram schematically illustrating a testing method for monitoring the blind spots in accordance with another example embodiment of the present invention.
  • FIG. 1 is a drawing schematically illustrating a learning device 100 for monitoring blind spots of a monitoring vehicle in accordance with one example embodiment of the present invention and, by referring to FIG. 1 , the learning device 100 may include a communication part 110 and a processor 120 .
  • the communication part 110 may receive training data corresponding to at least one output signal from a detector on the monitoring vehicle. That is, the communication part 110 may receive the output signal from the detector in following detailed description as an input signal, i.e., the training data.
  • the training data as information on monitored vehicles in an image corresponding to an input video from the detector, may include class information on objects such as the monitored vehicles and location information on areas where the monitored vehicles are located in the image. Additionally, the training data may be stored in a database 130 where GT, i.e., ground truth, of the class information and the location information on each of the monitored vehicles corresponding to the training data may be stored.
  • GT i.e., ground truth
  • the processor 120 may perform a first process of instructing a cue information extracting layer to perform one or more operations by using the class information and the location information on a monitored vehicle included in the training data, to thereby output one or more pieces of cue information on the monitored vehicle, a second process of instructing an FC layer for monitoring the blind spots to perform one or more neural network operations by using said pieces of cue information on the monitored vehicle, to thereby output a result of determining whether the monitored vehicle is located on one of the blind spots of the monitoring vehicle, and a third process of instructing a loss layer to generate one or more loss values by referring to the result and its corresponding GT, to thereby learn one or more parameters of the FC layer for monitoring the blind spots by backpropagating the loss values.
  • the learning device 100 in accordance with one example embodiment of the present invention may be a computing device and may include any device having a processor capable of computation in accordance with the present invention.
  • FIG. 1 represents one learning device 100 but the scope of the present invention is not limited thereto. That is, the learning device 100 may be configured as several devices to perform its functions.
  • a learning method for monitoring the blind spot of the monitoring vehicle by using the learning device configured as aforementioned in accordance with one example embodiment of the present invention may be explained as below, by referring to FIG. 2 .
  • the learning device 100 may instruct the cue information extracting layer 121 to perform the operations by using the class information and the location information on the monitored vehicle included in the training data, to thereby output said pieces of cue information on the monitored vehicle.
  • the detector capable of detecting the monitored vehicle in an image acquired by a vision sensor may be configured with a hypothesis generation stage of detecting an ROI (Region Of Interest), i.e., an area where the monitored vehicle is presumed to be located in the image, and a hypothesis verification stage of determining whether the detected ROI actually includes the monitored vehicle.
  • the hypothesis generation stage may be implemented with a motion-based scheme using optical flows, an appearance-based scheme using shadow underneath vehicles, corners, vertical and horizontal edges, symmetry, color, vehicle's lights, stereo cameras, multiple features, or the like
  • the hypothesis verification stage may be implemented with a correlation-based scheme using template matching, a learning-based scheme using features and classifiers, or the like.
  • the learning-based scheme may adopt a deep learning based algorithm such as a CNN, or a shallow learning based algorithm such as a decision tree, an SVM (Support Vector Machine), an AdaBoost, k-nearest neighbors, etc.
  • said pieces of cue information on the monitored vehicle may include at least part of (i) the class information on the monitored vehicle, (ii) the location information on the monitored vehicle, (iii) size information on the monitored vehicle corresponding to an ROI size, (iv) aspect ratio information on the monitored vehicle, and (v) distance information between a center of the monitored vehicle and one of outer sides of the blind spots.
  • the class information on the monitored vehicle may include class information for classifying the monitored vehicle as a car, a motorcycle, etc.
  • the location information on the monitored vehicle may include information on a location corresponding to an area where the monitored vehicle is located in the image.
  • coordinates of the top-left and the bottom-right corners of a bounding box in the image may be included, as information on a location of the bounding box corresponding to the monitored vehicle.
  • Coordinates of a center of the bounding box which may be calculated by using the coordinates of the top-left and the bottom-right corners of the bounding box, may further be included.
  • the distance information L may further include relative location information, e.g., information on plus sign or negative sign, to determine whether the center of the monitored vehicle is located inside or outside of said one of outer sides of the blind spots.
  • the learning device 100 may instruct the FC layer 122 for monitoring the blind spots to perform the neural network operations by using said pieces of the cue information on the monitored vehicle, to thereby output the result of determining whether the monitored vehicle is located on said one of the blind spots of the monitoring vehicle.
  • the FC layer 122 for monitoring the blind spots may include a neural network capable of outputting a result value of whether the monitored vehicle is located on said one of the blind spots by a multilayer perceptron to which said pieces of cue information on the monitored vehicle are inputted.
  • the number of inputs is represented as four, i.e., input 1 , input 2 , input 3 , and input 4 , for convenience of explanation, but the number of the inputs may be determined as corresponding to said pieces of cue information generated by the cue information extracting layer 121 .
  • a hidden layer is represented by one layer in FIG. 5 , the scope of the present invention is not limited thereto, and the hidden layer may include multiple layers.
  • the learning device 100 may instruct the loss layer 123 to generate the loss values by referring to the result from the FC layer 122 for monitoring the blind spots and its corresponding GT, to thereby learn the parameters of the FC layer 122 by backpropagating the loss values.
  • the cue information extracting layer 121 , the FC layer 122 , and the loss layer 123 may be included in one computing device or in different computing devices, and may also be implemented as an algorithm performing the operations previously mentioned, in any computing devices.
  • FIG. 6 is a drawing schematically illustrating a testing device 200 for monitoring blind spots of a monitoring vehicle, which may or may not be the same as the monitoring vehicle mentioned previously, in accordance with one example embodiment of the present invention, and the testing device 200 may include a communication part 210 and a processor 220 by referring to FIG. 6 .
  • the communication part 210 may receive class information for testing and location information for testing on a monitored vehicle from a detector 20 which detects the monitored vehicle in a test image taken from the monitoring vehicle.
  • the processor 220 may perform a first process of instructing the cue information extracting layer to perform the operations by using the class information for testing and the location information for testing on the monitored vehicle, to thereby output one or more pieces of cue information for testing on the monitored vehicle, and a second process of instructing the FC layer for monitoring the blind spots to perform the neural network operations by using said pieces of cue information for testing or their processed values on the monitored vehicle, to thereby output a result for testing of whether the monitored vehicle is located on one of the blind spots of the monitoring vehicle.
  • the FC layer for monitoring the blind spots has learned one or more parameters of the learning device using the learning method explained by referring to FIGS. 1 to 5 , and the learning method may be briefly explained as follows.
  • the learning device (i) if the training data corresponding to at least one output signal from the detector 20 on the monitoring vehicle has been inputted, has instructed the cue information extracting layer to perform the operations by using class information for training and location information for training on the monitored vehicle included in the training data, to thereby output one or more pieces of cue information for training on the monitored vehicle, (ii) has instructed the FC layer for monitoring the blind spots to perform the neural network operations by using said pieces of cue information for training on the monitored vehicle, to thereby output a result for training of determining whether the monitored vehicle is located on said one of the blind spots of the monitoring vehicle, and (iii) has instructed the loss layer to generate one or more loss values by referring to the result for training and its corresponding GT, to thereby learn the parameters of the FC layer for monitoring the blind spots by backpropagating the loss values
  • testing device 200 in accordance with one example embodiment of the present invention may be a computing device and may include any device having a processor capable of computation in accordance with the present invention.
  • FIG. 6 represents one testing device 200 but the scope of the present invention is not limited thereto. That is, the testing device 200 may be configured as several devices to perform its functions.
  • a testing method for monitoring the blind spots using the testing device for monitoring the blind spots configured as aforementioned in accordance with one example embodiment of the present invention may be explained as below, by referring to FIG. 7 .
  • the detector 20 may detect the monitored vehicle in the test image taken from the monitoring vehicle and may output the class information for testing and the location information for testing on the monitored vehicle.
  • the detector 20 capable of detecting the monitored vehicle in an image acquired by a vision sensor may be configured with a hypothesis generation stage of detecting an ROI, i.e., an area where the monitored vehicle is presumed to be located in the image, and a hypothesis verification stage of determining whether the detected ROI actually includes the monitored vehicle.
  • the hypothesis generation stage may be implemented with a motion-based scheme using optical flows, an appearance-based scheme using shadow underneath vehicles, corners, vertical and horizontal edges, symmetry, color, vehicle's lights, stereo cameras, multiple features, or the like
  • the hypothesis verification stage may be implemented with a correlation-based scheme using template matching, a learning-based scheme using features and classifiers, or the like.
  • the learning-based scheme may adopt a deep learning based algorithm, such as a CNN, or a shallow learning based algorithm, such as a decision tree, an SVM (Support Vector Machine), an AdaBoost, k-nearest neighbors, etc.
  • a deep learning based algorithm such as a CNN
  • a shallow learning based algorithm such as a decision tree, an SVM (Support Vector Machine), an AdaBoost, k-nearest neighbors, etc.
  • the detector 20 may be a vehicle detector based on an R-CNN (Region-based Convolutional Neural Network) including a convolutional layer 21 which generates a feature map for testing from the test image, an RPN (Region Proposal Network) 22 which generates an ROI for testing of the monitored vehicle from the feature map for testing, a pooling layer 23 which generates a feature vector for testing by pooling an area, in the feature map for testing, corresponding to the ROI for testing, at least one FC layer 24 for vehicle detection which performs at least one fully connected operation on the feature vector for testing, to thereby generate one or more FC output values, a classification layer 25 which outputs the class information for testing on the monitored vehicle by referring to the FC output values, and a regression layer 26 which outputs the location information for testing on the monitored vehicle by referring to the FC output values.
  • R-CNN Registered-based Convolutional Neural Network
  • each of the convolutional layer 21 and the FC layer 24 is presented as one layer, but the scope of the present invention is not limited thereto and each of them may be comprised of multiple layers.
  • the feature map for testing may be outputted in form of one or more feature maps corresponding to a depth of a channel, from the convolutional layer 21 .
  • the testing device 200 may instruct the cue information extracting layer 221 to perform the operations by using the class information for testing and the location information for testing on the monitored vehicle, to thereby output said pieces of cue information for testing on the monitored vehicle.
  • said pieces of cue information for testing on the monitored vehicle may include at least part of (i) the class information for testing on the monitored vehicle, (ii) the location information for testing on the monitored vehicle, (iii) size information for testing on the monitored vehicle corresponding to an ROI size for testing, (iv) aspect ratio information for testing on the monitored vehicle, and (v) distance information for testing between a center of the monitored vehicle and one of outer sides of the blind spots.
  • the class information for testing on the monitored vehicle may include class information for classifying the monitored vehicle as a car, a motorcycle, etc.
  • the location information for testing on the monitored vehicle may include information on a location corresponding to an area where the monitored vehicle is located in the test image.
  • coordinates of the top-left and the bottom-right corners of a bounding box in the test image may be included, as information on a location of the bounding box corresponding to the monitored vehicle.
  • Coordinates of a center of the bounding box which may be calculated by using the coordinates of the top-left and the bottom-right corners of the bounding box, may be further included.
  • the distance information L for testing may include relative location information for testing, e.g., information on plus sign or negative sign, to determine whether the center of the monitored vehicle is located inside or outside of said one of outer sides of the blind spots.
  • the testing device 200 may instruct the FC layer 222 for monitoring the blind spots to perform the neural network operations by using said pieces of the cue information for testing or their processed values on the monitored vehicle, to thereby output a result for testing of determining whether the monitored vehicle is located on said one of the blind spots of the monitoring vehicle.
  • the processed values of said pieces of cue information for testing may represent a concatenated value acquired by using said pieces of cue information for testing on the monitored vehicle generated by the cue information extracting layer 221 and the feature vector for testing generated by the pooling layer 23 of the detector 20 .
  • a feature concatenation layer capable of concatenating said pieces of cue information for testing and the feature vector for testing may be located before the FC layer 222 for monitoring the blind spots.
  • the FC layer 222 may include the neural network capable of outputting a result value of whether the monitored vehicle is located on said one of the blind spots by the multilayer perceptron to which said pieces of cue information for testing or their processed values on the monitored vehicle are inputted.
  • the cue information extracting layer 221 and the FC layer 222 for monitoring the blind spots may be included in one computing device or in different computing devices, and may also be implemented as an algorithm performing the operations previously mentioned, in any computing devices.
  • FIG. 9 is a drawing schematically illustrating a learning device 300 for monitoring blind spots of a monitoring vehicle in accordance with another example embodiment of the present invention, and the learning device 300 may include a communication part 310 and a processor 320 by referring to FIG. 9 .
  • the communication part 310 may receive a training image corresponding to at least one video image taken from the monitoring vehicle.
  • the training image may be stored in a database 330 where at least one GT of class information and location information on a monitored vehicle corresponding to the training image may be stored.
  • the processor 320 may perform a first process of instructing a detector 30 on the monitoring vehicle to output the class information and the location information on the monitored vehicle, a second process of instructing a cue information extracting layer to perform one or more operations by using the class information and the location information on the monitored vehicle, to thereby output one or more pieces of cue information on the monitored vehicle, and instructing an FC layer for monitoring the blind spots to perform one or more neural network operations by using said pieces of cue information or their processed values on the monitored vehicle, to thereby output a result of determining whether the monitored vehicle is located on one of the blind spots, and a third process of instructing a first loss layer to generate one or more loss values for the blind spots by referring to the result and its corresponding first GT, to thereby learn one or more parameters of the FC layer for monitoring the blind spots by backpropagating the loss values for the blind spots, and instructing a second loss layer to generate one or more loss values for vehicle detection by referring to the class information and the location information on the monitored vehicle and their corresponding second GT,
  • the learning device 300 in accordance with another example embodiment of the present invention may be a computing device and may include any device having a processor capable of computation in accordance with the present invention.
  • FIG. 9 represents one learning device 300 but the scope of the present invention is not limited thereto. That is, the learning device 300 may be configured as several devices to perform its functions.
  • a learning method for monitoring the blind spot of the monitoring vehicle by using the learning device configured as aforementioned in accordance with another example embodiment of the present invention may be explained as below, by referring to FIG. 10 .
  • the detector 30 may output the class information and the location information on the monitored vehicle included in the training image.
  • the detector 30 capable of detecting the monitored vehicle in an image acquired by a vision sensor may be configured with a hypothesis generation stage of detecting an ROI, i.e., an area where the monitored vehicle is presumed to be located in the image, and a hypothesis verification stage of determining whether the detected ROI actually includes the monitored vehicle.
  • the hypothesis generation stage may be implemented with a motion-based scheme using optical flows, an appearance-based scheme using shadow underneath vehicles, corners, vertical and horizontal edges, symmetry, color, vehicle's lights, stereo cameras, multiple features, or the like
  • the hypothesis verification stage may be implemented with a correlation-based scheme using template matching, a learning-based scheme using features and classifiers, or the like.
  • the learning-based scheme may adopt a deep learning based algorithm, such as a CNN, or a shallow learning based algorithm, such as a decision tree, an SVM (Support Vector Machine), an AdaBoost, k-nearest neighbors, etc.
  • a deep learning based algorithm such as a CNN
  • a shallow learning based algorithm such as a decision tree, an SVM (Support Vector Machine), an AdaBoost, k-nearest neighbors, etc.
  • the detector 30 may be a vehicle detector based on an R-CNN, including a convolutional layer 31 which generates a feature map from the training image, an RPN 32 which generates the ROI of the monitored vehicle from the feature map, a pooling layer 33 which generates a feature vector by pooling an area, in the feature map, corresponding to the ROI, and at least one FC layer 34 for the vehicle detection which performs at least one fully connected operation on the feature vector to thereby generate one or more FC output values, a classification layer 35 which outputs the class information on the monitored vehicle by referring to the FC output values, and a regression layer 36 which outputs the location information on the monitored vehicle by referring to the FC output values.
  • the convolutional layer 31 may be comprised of at least one layer and the FC layer 34 may also be comprised of at least one layer.
  • the learning device 300 may instruct the cue information extracting layer 321 to perform the operations by using the class information and the location information on the monitored vehicle, to thereby output said pieces of cue information on the monitored vehicle.
  • said pieces of cue information on the monitored vehicle may include at least part of (i) the class information on the monitored vehicle, (ii) the location information on the monitored vehicle, (iii) size information on the monitored vehicle corresponding to an ROI size, (iv) aspect ratio information on the monitored vehicle, and (v) distance information between a center of the monitored vehicle and one of outer sides of the blind spots.
  • the class information on the monitored vehicle may include class information for classifying the monitored vehicle as a car, a motorcycle, etc.
  • the location information on the monitored vehicle may include information on a location corresponding to an area where the monitored vehicle is located in the image. For example, coordinates of the top-left and the bottom-right corners of a bounding box in the image may be included, as information on a location of the bounding box corresponding to the monitored vehicle. Coordinates of a center of the bounding box, which may be calculated by using the coordinates of the top-left and the bottom-right corners of the bounding box, may be further included.
  • the distance information L may include relative location information, e.g., information on plus sign or negative sign, to determine whether the center of the monitored vehicle is located inside or outside of said one of outer sides of the blind spots.
  • the learning device 300 may instruct the FC layer 322 for monitoring the blind spots to perform the neural network operations by using said pieces of the cue information or their processed values on the monitored vehicle, to thereby output the result of determining whether the monitored vehicle is located on said one of the blind spots of the monitoring vehicle.
  • the processed values of said pieces of cue information may represent a concatenated value acquired by using said pieces of cue information on the monitored vehicle generated by the cue information extracting layer 321 and the feature vector generated by the pooling layer 33 of the detector 30 .
  • a feature concatenation layer 340 which generates the concatenated value including said pieces of cue information and the feature vector, may be located before the FC layer 322 for monitoring the blind spots.
  • the FC layer 322 may include a neural network capable of outputting a result value of whether the monitored vehicle is located on said one of the blind spots by a multilayer perceptron to which said pieces of cue information or their processed values on the monitored vehicle are inputted.
  • the learning device 300 may instruct the first loss layer 323 to generate the loss values for the blind spots by referring to (i) the result of determining whether the monitored vehicle is located on said one of the blind spots of the monitoring vehicle and (ii) its corresponding first GT, to thereby learn the parameters of the FC layer 322 for monitoring the blind spots by backpropagating the loss values for the blind spots, and instructing the second loss layer 324 to generate the loss values for the vehicle detection by referring to the class information and the location information on the monitored vehicle and their corresponding second GT, to thereby learn the parameters of the detector 30 by backpropagating the loss values for the vehicle detection.
  • the loss values for the vehicle detection may include one or more class loss values and one or more location loss values of the monitored vehicle
  • the learning device 300 may learn one or more parameters of the FC layer 34 for the vehicle detection and one or more parameters of the convolutional layer 31 of the detector 30 by backpropagating the class loss values and the location loss values.
  • the learning device 300 may acquire one or more ROI loss values by referring to the ROI, which is generated from the RPN 32 of the detector 30 , and ROI GT corresponding to the training image, to thereby learn one or more parameters of the RPN 32 by backpropagating the ROI loss values.
  • the cue information extracting layer 221 and the FC layer 222 for monitoring the blind spots may be included in one computing device or in different computing devices, and may also be implemented as an algorithm performing the operations previously mentioned, in any computing devices. Additionally, the first loss layer 323 and the second loss layer 324 may be included in one computing device or in different computing devices respectively and may be implemented as an algorithm performing the operations previously mentioned, in any computing devices.
  • FIG. 11 is a drawing schematically illustrating a testing device 400 for monitoring blind spots of a monitoring vehicle, which may or may not be the same as the monitoring vehicle mentioned previously, in accordance with another example embodiment of the present invention, and the testing device 400 may include a communication part 410 and a processor 420 by referring to FIG. 11 .
  • the communication part 410 may receive class information for testing and location information for testing on a monitored vehicle from a detector 40 which detects the monitored vehicle in a test image taken from the monitoring vehicle.
  • the detector 40 capable of detecting the monitored vehicle in an image acquired by a vision sensor may be configured with a hypothesis generation stage of detecting an ROI, i.e., an area where the monitored vehicle is presumed to be located in the image and a hypothesis verification stage of determining whether the detected ROI actually includes the monitored vehicle.
  • the hypothesis generation stage may be implemented with a motion-based scheme using optical flows, an appearance-based scheme using shadow underneath vehicles, corners, vertical and horizontal edges, symmetry, color, vehicle's lights, stereo cameras, multiple features, or the like
  • the hypothesis verification stage may be implemented with a correlation-based scheme using template matching, a learning-based scheme using features and classifiers, or the like.
  • the learning-based scheme may adopt a deep learning based algorithm, such as a CNN, or a shallow learning based algorithm, such as a decision tree, an SVM (Support Vector Machine), an AdaBoost, k-nearest neighbors, etc.
  • a deep learning based algorithm such as a CNN
  • a shallow learning based algorithm such as a decision tree, an SVM (Support Vector Machine), an AdaBoost, k-nearest neighbors, etc.
  • the detector 40 may be a vehicle detector based on the R-CNN, including the convolutional layer which generates a feature map for testing from the test image, the RPN which generates an ROI for testing of the monitored vehicle from the feature map for testing, the pooling layer which generates a feature vector for testing by pooling an area, in the feature map for testing, corresponding to the ROI for testing, the FC layer for the vehicle detection which performs the fully connected operations by using the feature vector for testing, to thereby generate one or more FC output values, the classification layer which outputs the class information for testing on the monitored vehicle by referring to the FC output values, and the regression layer which outputs the location information for testing on the monitored vehicle by referring to the FC output values.
  • the convolutional layer which generates a feature map for testing from the test image
  • the RPN which generates an ROI for testing of the monitored vehicle from the feature map for testing
  • the pooling layer which generates a feature vector for testing by pooling an area, in the feature map for testing, corresponding to the ROI for testing
  • each of the convolutional layer and the FC layer for the vehicle detection is presented as one layer, but the scope of the present invention is not limited thereto and each of them may be comprised of multiple layers.
  • the feature map for testing may be outputted in form of one or more feature maps corresponding to a depth of a channel, from the convolutional layer 41 .
  • the processor 420 may perform a first process of instructing the cue information extracting layer to perform the operations by using the class information for testing and the location information for testing on the monitored vehicle, to thereby output one or more pieces of cue information for testing on the monitored vehicle, and a second process of instructing the FC layer for monitoring the blind spots to perform the neural network operations by using said pieces of cue information for testing or their processed values on the monitored vehicle, to thereby output a result for testing of whether the monitored vehicle is located on one of the blind spots of the monitoring vehicle.
  • the detector 40 and the FC layer for monitoring the blind spots have learned the parameters of the learning device using the learning method explained by referring to FIGS. 9 and 10 , and the learning method may be briefly explained as follows. If the detector 40 has outputted class information for training and location information for training on the monitored vehicle in a training image corresponding to a video image taken from the monitoring vehicle, the learning device (i) has instructed the cue information extracting layer to perform the operations by using the class information for training and the location information for training on the monitored vehicle, to thereby output one or more pieces of cue information for training on the monitored vehicle, (ii) has instructed the FC layer for monitoring the blind spots to perform the neural network operations by using said pieces of cue information for training or their processed values, to thereby output a result for training of whether the monitored vehicle is located on said one of the blind spots, and (iii) has instructed the first loss layer to generate one or more loss values for the blind spots by referring to the result for training and its corresponding first GT, to thereby learn the parameters of the FC layer for monitoring the
  • testing device 400 in accordance with another example embodiment of the present invention may be a computing device and may include any device having a processor capable of computation in accordance with the present invention.
  • FIG. 11 represents one testing device 400 but the scope of the present invention is not limited thereto. That is, the testing device 400 may be configured as several devices to perform its functions.
  • a testing method for monitoring the blind spots using the testing device for monitoring the blind spots configured as aforementioned in accordance with another example embodiment of the present invention may be explained as below, by referring to FIG. 12 .
  • the learning device (i) has instructed the cue information extracting layer 421 to perform the operations by using the class information for training and location information for training on the monitored vehicle, to thereby output said pieces of cue information for training on the monitored vehicle, (ii) has instructed the FC layer 422 for monitoring the blind spots to perform the neural network operations by using said pieces of cue information for training or their processed values, to thereby output the result for training of whether the monitored vehicle is located on said one of the blind spots, and (iii) has instructed the first loss layer to generate the loss values for the blind spots by referring to the result for training and its corresponding first GT, to thereby learn the parameters of the FC layer 422 for monitoring the blind spots by backpropagating the loss values for the blind spots, and has instructed the second loss layer to generate the loss values for the vehicle detection by referring to the class information for training and the location
  • the detector 40 capable of detecting the monitored vehicle in an image acquired by the vision sensor may be configured with the hypothesis generation stage of detecting the ROI, i.e., the area where the monitored vehicle is presumed to be located in the image, and the hypothesis verification stage of determining whether the detected ROI actually includes the monitored vehicle.
  • the hypothesis generation stage may be implemented with a motion-based scheme using optical flows, an appearance-based scheme using shadow underneath vehicle, corners, vertical and horizontal edges, symmetry, color, vehicle's lights, stereo cameras, multiple features, or the like
  • the hypothesis verification stage may be implemented with a correlation-based scheme using template matching, a learning-based scheme using features and classifiers, or the like.
  • the learning-based scheme may adopt a deep learning based algorithm, such as a CNN, or a shallow learning based algorithm, such as a decision tree, an SVM (Support Vector Machine), an AdaBoost, k-nearest neighbors, etc.
  • a deep learning based algorithm such as a CNN
  • a shallow learning based algorithm such as a decision tree, an SVM (Support Vector Machine), an AdaBoost, k-nearest neighbors, etc.
  • the detector 40 may be a vehicle detector based on the R-CNN, including the convolutional layer 41 which generates a feature map for testing from the test image, the RPN 42 which generates the ROI for testing of the monitored vehicle from the feature map for testing, the pooling layer 43 which generates a feature vector for testing by pooling an area, in the feature map for testing, corresponding to the ROI for testing, the FC layer 44 for the vehicle detection which performs the fully connected operation on the feature vector for testing, to thereby generate one or more FC output values, the classification layer 45 which outputs the class information for testing on the monitored vehicle by referring to the FC output values, and the regression layer 46 which outputs the location information for testing on the monitored vehicle by referring to the FC output values.
  • each of the convolutional layer 41 and the FC layer 44 for the vehicle detection is presented as one layer, but the scope of the present invention is not limited thereto and each of them may be comprised of multiple layers.
  • the testing device 400 may instruct the cue information extracting layer 421 to perform the operations by using the class information for testing and the location information for testing on the monitored vehicle, to thereby output said pieces of cue information for testing on the monitored vehicle.
  • said pieces of cue information for testing on the monitored vehicle may include at least part of (i) the class information for testing on the monitored vehicle, (ii) the location information for testing on the monitored vehicle, (iii) size information for testing on the monitored vehicle corresponding to an ROI size for testing, (iv) aspect ratio information for testing on the monitored vehicle, and (v) distance information for testing between a center of the monitored vehicle and one of outer sides of the blind spots.
  • the class information for testing on the monitored vehicle may include class information for classifying the monitored vehicle as a car, a motorcycle, etc.
  • the location information for testing on the monitored vehicle may include information on a location corresponding to an area where the monitored vehicle is located in the test image. For example, coordinates of the top-left and the bottom-right corners of a bounding box in the test image may be included, as information on a location of the bounding box corresponding to the monitored vehicle. Coordinates of a center of the bounding box, which may be calculated by using the coordinates of the top-left and the bottom-right corners of the bounding box, may be further included.
  • the distance information L for testing may include relative location information for testing, e.g., information on plus sign or negative sign, to determine whether the center of the monitored vehicle is located inside or outside of said one of outer sides of the blind spots.
  • the testing device 400 may instruct the FC layer 422 for monitoring the blind spots to perform the neural network operations by using said pieces of the cue information for testing or their processed values on the monitored vehicle, to thereby output the result for testing of determining whether the monitored vehicle is located on said one of the blind spots of the monitoring vehicle.
  • the processed values of said pieces of cue information for testing may represent a concatenated value acquired by using said pieces of cue information for testing on the monitored vehicle generated by the cue information extracting layer 421 and the feature vector for testing generated by the pooling layer 43 of the detector 40 .
  • a feature concatenation layer 440 which generates the concatenated value including said pieces of cue information for testing and the feature vector for testing, may be located before the FC layer 422 for monitoring the blind spots.
  • the FC layer 422 for monitoring the blind spots may include the neural network capable of outputting a result value of whether the monitored vehicle is located on said one of the blind spots by the multilayer perceptron to which said pieces of cue information for testing or their processed values on the monitored vehicle are inputted.
  • the cue information extracting layer 421 and the FC layer 422 for monitoring the blind spots may be included in one computing device or in different computing devices, and also may be implemented as an algorithm performing the operations previously mentioned, in any computing devices.
  • the present invention has an effect of providing a blind spot monitoring system regardless of the type of a detector which detects a vehicle.
  • the present invention has another effect of determining whether the vehicle is located on blind spots by using an output signal from the detector, regardless of the type of the detector.
  • the present invention has still another effect of minimizing cost of maintaining the blind spot monitoring system since the detector in the blind spot monitoring system is replaceable as needed without designing a logic.
  • the present invention has still yet another effect of no need to design the logic corresponding to each detector for determining whether the vehicle is located on the blind spot, thereby minimizing time required to develop the blind spot monitoring system, as the blind spot monitoring system is adaptable to all detectors.
  • the embodiments of the present invention as explained above can be implemented in a form of executable program command through a variety of computer means recordable to computer readable media.
  • the computer readable media may include solely or in combination, program commands, data files, and data structures.
  • the program commands recorded to the media may be components specially designed for the present invention or may be usable to a skilled human in a field of computer software.
  • Computer readable media include magnetic media such as hard disk, floppy disk, and magnetic tape, optical media such as CD-ROM and DVD, magneto-optical media such as floptical disk and hardware devices such as ROM, RAM, and flash memory specially designed to store and carry out program commands.
  • Program commands include not only a machine language code made by a compiler but also a high level language code that can be executed by a computer using an interpreter, etc.
  • the hardware device can work as more than a software module to perform the process in accordance with the present invention and they can do the same in the opposite case.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Multimedia (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Biodiversity & Conservation Biology (AREA)
  • Automation & Control Theory (AREA)
  • Transportation (AREA)
  • Mechanical Engineering (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Image Analysis (AREA)
  • Traffic Control Systems (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
US16/152,699 2018-10-05 2018-10-05 Learning method and testing method for monitoring blind spot of vehicle, and learning device and testing device using the same Active US10474930B1 (en)

Priority Applications (5)

Application Number Priority Date Filing Date Title
US16/152,699 US10474930B1 (en) 2018-10-05 2018-10-05 Learning method and testing method for monitoring blind spot of vehicle, and learning device and testing device using the same
KR1020190105496A KR102349953B1 (ko) 2018-10-05 2019-08-27 자동차의 블라인드 스팟 모니터링을 위한 학습 방법과 테스팅 방법, 및 이를 이용한 학습 장치와 테스팅 장치
EP19195517.8A EP3633558A1 (en) 2018-10-05 2019-09-05 Learning method and testing method for monitoring blind spot of vehicle, and learning device and testing device using the same
JP2019182005A JP6814334B2 (ja) 2018-10-05 2019-10-02 ブラインドスポットモニタリングのためのｃｎｎの学習方法、テスティング方法、学習装置、及びテスティング装置
CN201910949146.3A CN111008553B (zh) 2018-10-05 2019-10-08 用于监测车辆盲区的方法和装置

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
US16/152,699 US10474930B1 (en) 2018-10-05 2018-10-05 Learning method and testing method for monitoring blind spot of vehicle, and learning device and testing device using the same

Publications (1)

Publication Number Publication Date
US10474930B1 true US10474930B1 (en) 2019-11-12

Family

ID=67874282

Family Applications (1)

Application Number Title Priority Date Filing Date
US16/152,699 Active US10474930B1 (en) 2018-10-05 2018-10-05 Learning method and testing method for monitoring blind spot of vehicle, and learning device and testing device using the same

Country Status (5)

Country Link
US (1) US10474930B1 (ko)
EP (1) EP3633558A1 (ko)
JP (1) JP6814334B2 (ko)
KR (1) KR102349953B1 (ko)
CN (1) CN111008553B (ko)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10726303B1 (en) * 2019-01-30 2020-07-28 StradVision, Inc. Learning method and learning device for switching modes of autonomous vehicle based on on-device standalone prediction to thereby achieve safety of autonomous driving, and testing method and testing device using the same
US20200250514A1 (en) * 2019-01-31 2020-08-06 StradVision, Inc. Learning method and learning device for integrating object detection information acquired through v2v communication from other autonomous vehicle with object detection information generated by present autonomous vehicle, and testing method and testing device using the same
CN111652128A (zh) * 2020-06-02 2020-09-11 浙江大华技术股份有限公司 一种高空电力作业安全监测方法、系统和存储装置
US10853698B2 (en) * 2016-11-09 2020-12-01 Konica Minolta Laboratory U.S.A., Inc. System and method of using multi-frame image features for object detection

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113239912B (zh) * 2021-07-13 2021-09-17 天津所托瑞安汽车科技有限公司 Bsd图像有效区域的确定方法、设备和存储介质
KR102496765B1 (ko) 2022-10-11 2023-02-07 유희경 자동차용 블라인드

Citations (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6911642B2 (en) * 2000-02-18 2005-06-28 Ficosa North America Corporation Object presence detection method and device
US20110106370A1 (en) * 2006-03-14 2011-05-05 Airmax Group Plc Method and system for driver style monitoring and analysing
US8346480B2 (en) * 2006-03-16 2013-01-01 Gray & Company, Inc. Navigation and control system for autonomous vehicles
US20150258990A1 (en) * 2013-05-24 2015-09-17 Advanced Scientific Concepts, Inc. Automotive auxiliary ladar sensor
US20170103340A1 (en) * 2015-10-09 2017-04-13 Fair Isaac Corporation Method for Real-Time Enhancement of a Predictive Algorithm by a Novel Measurement of Concept Drift Using Algorithmically-Generated Features
US9805274B2 (en) * 2016-02-03 2017-10-31 Honda Motor Co., Ltd. Partially occluded object detection using context and depth ordering
US20170357859A1 (en) * 2016-06-13 2017-12-14 Ford Global Technologies, Llc Blind Spot Detection Systems And Methods
US20170355263A1 (en) * 2016-06-13 2017-12-14 Ford Global Technologies, Llc Blind Spot Detection Systems And Methods
US9873196B2 (en) * 2015-06-24 2018-01-23 Brain Corporation Bistatic object detection apparatus and methods
US9934440B1 (en) * 2017-10-04 2018-04-03 StradVision, Inc. Method for monitoring blind spot of monitoring vehicle and blind spot monitor using the same
US9947228B1 (en) * 2017-10-05 2018-04-17 StradVision, Inc. Method for monitoring blind spot of vehicle and blind spot monitor using the same
US10099614B2 (en) * 2011-11-28 2018-10-16 Magna Electronics Inc. Vision system for vehicle
US10106153B1 (en) * 2018-03-23 2018-10-23 Chongqing Jinkang New Energy Vehicle Co., Ltd. Multi-network-based path generation for vehicle parking

Family Cites Families (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH10181603A (ja) * 1996-12-25 1998-07-07 Toshiba Corp 乗車列車選択支援情報作成装置
JP4552305B2 (ja) * 2000-10-02 2010-09-29 日産自動車株式会社 画像表示装置および表示方法
ES2177469B1 (es) * 2001-05-16 2003-12-16 Fico Mirrors Sa Dispositivo de deteccion de presencia de objetos en un angulo muerto de un vehiculo automovil.
US6859148B2 (en) * 2002-10-30 2005-02-22 Ford Global Technologies, Llc Blind spot warning system for an automotive vehicle
JP5620147B2 (ja) * 2010-05-24 2014-11-05 株式会社豊田中央研究所 可動物予測装置及びプログラム
JP2016006626A (ja) * 2014-05-28 2016-01-14 株式会社デンソーアイティーラボラトリ 検知装置、検知プログラム、検知方法、車両、パラメータ算出装置、パラメータ算出プログラムおよびパラメータ算出方法
KR101692628B1 (ko) * 2014-12-24 2017-01-04 한동대학교 산학협력단 관심영역을 이용하여 차량의 후방 좌우 옆 차선 영역을 감지하는 방법 및 이를 이용한 차량용 영상 모니터링 시스템
US10395764B2 (en) * 2015-01-06 2019-08-27 Aic Innovations Group, Inc. Method and apparatus for recognition of patient activity
US20170124409A1 (en) * 2015-11-04 2017-05-04 Nec Laboratories America, Inc. Cascaded neural network with scale dependent pooling for object detection
US10002313B2 (en) * 2015-12-15 2018-06-19 Sighthound, Inc. Deeply learned convolutional neural networks (CNNS) for object localization and classification
JP2017162438A (ja) * 2016-03-11 2017-09-14 パナソニック インテレクチュアル プロパティ コーポレーション オブ アメリカＰａｎａｓｏｎｉｃ Ｉｎｔｅｌｌｅｃｔｕａｌ Ｐｒｏｐｅｒｔｙ Ｃｏｒｐｏｒａｔｉｏｎ ｏｆ Ａｍｅｒｉｃａ 危険予測方法
US10055652B2 (en) * 2016-03-21 2018-08-21 Ford Global Technologies, Llc Pedestrian detection and motion prediction with rear-facing camera
CN106407931B (zh) * 2016-09-19 2019-11-22 杭州电子科技大学 一种深度卷积神经网络运动车辆检测方法
CN106599939A (zh) * 2016-12-30 2017-04-26 深圳市唯特视科技有限公司 一种基于区域卷积神经网络的实时目标检测方法
JP6719399B2 (ja) * 2017-02-10 2020-07-08 ヤフー株式会社 解析装置、解析方法、およびプログラム
JP6823495B2 (ja) * 2017-02-27 2021-02-03 株式会社日立製作所 情報処理装置および画像認識装置
US11308391B2 (en) * 2017-03-06 2022-04-19 Baidu Usa Llc Offline combination of convolutional/deconvolutional and batch-norm layers of convolutional neural network models for autonomous driving vehicles
US10083375B1 (en) * 2017-10-13 2018-09-25 StradVision, Inc. Method and device for performing activation and convolution operation at the same time and learning method and learning device for the same

Patent Citations (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6911642B2 (en) * 2000-02-18 2005-06-28 Ficosa North America Corporation Object presence detection method and device
US20110106370A1 (en) * 2006-03-14 2011-05-05 Airmax Group Plc Method and system for driver style monitoring and analysing
US8346480B2 (en) * 2006-03-16 2013-01-01 Gray & Company, Inc. Navigation and control system for autonomous vehicles
US10099614B2 (en) * 2011-11-28 2018-10-16 Magna Electronics Inc. Vision system for vehicle
US20150258990A1 (en) * 2013-05-24 2015-09-17 Advanced Scientific Concepts, Inc. Automotive auxiliary ladar sensor
US9873196B2 (en) * 2015-06-24 2018-01-23 Brain Corporation Bistatic object detection apparatus and methods
US20170103340A1 (en) * 2015-10-09 2017-04-13 Fair Isaac Corporation Method for Real-Time Enhancement of a Predictive Algorithm by a Novel Measurement of Concept Drift Using Algorithmically-Generated Features
US9805274B2 (en) * 2016-02-03 2017-10-31 Honda Motor Co., Ltd. Partially occluded object detection using context and depth ordering
US20170357859A1 (en) * 2016-06-13 2017-12-14 Ford Global Technologies, Llc Blind Spot Detection Systems And Methods
US20170355263A1 (en) * 2016-06-13 2017-12-14 Ford Global Technologies, Llc Blind Spot Detection Systems And Methods
US9934440B1 (en) * 2017-10-04 2018-04-03 StradVision, Inc. Method for monitoring blind spot of monitoring vehicle and blind spot monitor using the same
US9947228B1 (en) * 2017-10-05 2018-04-17 StradVision, Inc. Method for monitoring blind spot of vehicle and blind spot monitor using the same
US10106153B1 (en) * 2018-03-23 2018-10-23 Chongqing Jinkang New Energy Vehicle Co., Ltd. Multi-network-based path generation for vehicle parking

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10853698B2 (en) * 2016-11-09 2020-12-01 Konica Minolta Laboratory U.S.A., Inc. System and method of using multi-frame image features for object detection
US10726303B1 (en) * 2019-01-30 2020-07-28 StradVision, Inc. Learning method and learning device for switching modes of autonomous vehicle based on on-device standalone prediction to thereby achieve safety of autonomous driving, and testing method and testing device using the same
US20200250514A1 (en) * 2019-01-31 2020-08-06 StradVision, Inc. Learning method and learning device for integrating object detection information acquired through v2v communication from other autonomous vehicle with object detection information generated by present autonomous vehicle, and testing method and testing device using the same
US10817777B2 (en) * 2019-01-31 2020-10-27 StradVision, Inc. Learning method and learning device for integrating object detection information acquired through V2V communication from other autonomous vehicle with object detection information generated by present autonomous vehicle, and testing method and testing device using the same
CN111652128A (zh) * 2020-06-02 2020-09-11 浙江大华技术股份有限公司 一种高空电力作业安全监测方法、系统和存储装置
CN111652128B (zh) * 2020-06-02 2023-09-01 浙江大华技术股份有限公司 一种高空电力作业安全监测方法、系统和存储装置

Also Published As

Publication number Publication date
CN111008553A (zh) 2020-04-14
CN111008553B (zh) 2023-11-28
KR102349953B1 (ko) 2022-01-12
KR20200039548A (ko) 2020-04-16
JP6814334B2 (ja) 2021-01-20
JP2020061140A (ja) 2020-04-16
EP3633558A1 (en) 2020-04-08

Similar Documents

Publication Publication Date Title
US10474930B1 (en) Learning method and testing method for monitoring blind spot of vehicle, and learning device and testing device using the same
US9934440B1 (en) Method for monitoring blind spot of monitoring vehicle and blind spot monitor using the same
US10643085B1 (en) Method and device for estimating height and weight of passengers using body part length and face information based on human's status recognition
Satzoda et al. Multipart vehicle detection using symmetry-derived analysis and active learning
US10300851B1 (en) Method for warning vehicle of risk of lane change and alarm device using the same
US20180204076A1 (en) Moving object detection and classification image analysis methods and systems
CN102997900B (zh) 外界识别方法、装置以及车辆系统
JP2019075116A (ja) トラッキングネットワークを含むＣＮＮ（Ｃｏｎｖｏｌｕｔｉｏｎａｌ Ｎｅｕｒａｌ Ｎｅｔｗｏｒｋ）を用いてイメージ上の客体に対応するバウンディングボックスを獲得するための方法及びこれを利用した装置
CN107368890A (zh) 一种以视觉为中心的基于深度学习的路况分析方法及系统
US9928404B2 (en) Determination device, determination method, and non-transitory storage medium
US10984262B2 (en) Learning method and testing method for monitoring blind spot of vehicle, and learning device and testing device using the same
US9626599B2 (en) Reconfigurable clear path detection system
US10635917B1 (en) Method and device for detecting vehicle occupancy using passenger's keypoint detected through image analysis for humans' status recognition
JP7135665B2 (ja) 車両制御システム、車両の制御方法及びコンピュータプログラム
US10373323B1 (en) Method and device for merging object detection information detected by each of object detectors corresponding to each camera nearby for the purpose of collaborative driving by using V2X-enabled applications, sensor fusion via multiple vehicles
US11106214B2 (en) Artificial neural network-based projection information recognition apparatus and method thereof
US20220171975A1 (en) Method for Determining a Semantic Free Space
US20200174488A1 (en) False target removal device and method for vehicles and vehicle including the device
KR20210008260A (ko) 자율주행차량의 거동 제어 장치 및 그 방법
Zakaria et al. A novel vehicle detection system
KR102107298B1 (ko) 영상 분석 장치 및 방법
Garcia et al. Mobile based pedestrian detection with accurate tracking
Liu Development of a vision-based object detection and recognition system for intelligent vehicle
JP7379268B2 (ja) 画像処理装置
US20240001933A1 (en) Occupant temperature estimating device, occupant state detection device, occupant temperature estimating method, and occupant temperature estimating system

Legal Events

Date Code Title Description
FEPP Fee payment procedure

Free format text: ENTITY STATUS SET TO UNDISCOUNTED (ORIGINAL EVENT CODE: BIG.); ENTITY STATUS OF PATENT OWNER: SMALL ENTITY

FEPP Fee payment procedure

Free format text: ENTITY STATUS SET TO SMALL (ORIGINAL EVENT CODE: SMAL); ENTITY STATUS OF PATENT OWNER: SMALL ENTITY

STCF Information on status: patent grant

Free format text: PATENTED CASE

MAFP Maintenance fee payment

Free format text: PAYMENT OF MAINTENANCE FEE, 4TH YR, SMALL ENTITY (ORIGINAL EVENT CODE: M2551); ENTITY STATUS OF PATENT OWNER: SMALL ENTITY

Year of fee payment: 4